A probabilistic machine learning framework of flutter critical wind velocity prediction for cable-stayed bridges based on a calibrated finite element model
Transport and Communications Science Journal · 25.09.2026 · 77 (9) · 1185-1197
DOI: 10.47869/tcsj.77.9.5 ↗Kurz gefasst
Fachartikel in der Zeitschrift Transport and Communications Science Journal, online erschienen am 25.09.2026. Verfasst von Nam Ho Xuan und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Flutter is a self-excited aeroelastic instability that controls the wind design of long-span cable-stayed bridges. It is usually assessed by wind tunnel tests or costly simulations. This study examines whether a probabilistic machine learning model can give a first estimate of the flutter critical wind velocity, based on flat-plate theory, from preliminary design parameters, for a family of bridges derived from one existing bridge. A finite element model of the Dakrong cable-stayed bridge in Vietnam is updated to match ten of the twelve natural frequencies measured on site (root-mean-square error 4.7%). Three hundred bridge configurations are generated from the updated model by Latin Hypercube Sampling of eleven variables within slenderness limits, and their critical velocity is estimated with the Selberg formula and checked with the Rocard formula. On a test set of 61 configurations, Gaussian process regression with automatic relevance determination, trained only on span, width, depth, mass and inertia, reproduces the Selberg estimate with R2 = 0.92 and a mean absolute percentage error of 6.4%, and 95.1% of the test values fall inside its 95% confidence intervals. These values measure agreement with the formula-based labels, not with the actual flutter speed. The labels assume flat-plate aerodynamics and a single pair of modes, and additional checks suggest that, for this open-girder deck, their uncertainty is larger than the prediction error. The model is therefore intended only for comparing deck alternatives within the sampled family and does not replace a flutter check.
Zitieren
APA
Xuan, N. H., & Ngoc, L. N. (2026). A probabilistic machine learning framework of flutter critical wind velocity prediction for cable-stayed bridges based on a calibrated finite element model. Transport and Communications Science Journal, 77(9), 1185-1197. https://doi.org/10.47869/tcsj.77.9.5
Vancouver
Xuan NH, Ngoc LN. A probabilistic machine learning framework of flutter critical wind velocity prediction for cable-stayed bridges based on a calibrated finite element model. Transport and Communications Science Journal. 2026;77(9):1185-1197. doi:10.47869/tcsj.77.9.5
BibTeX
@article{Xuan2026,
author = {Xuan, Nam Ho and Ngoc, Lan Nguyen},
title = {A probabilistic machine learning framework of flutter critical wind velocity prediction for cable-stayed bridges based on a calibrated finite element model},
journal = {Transport and Communications Science Journal},
year = {2026},
volume = {77},
number = {9},
pages = {1185--1197},
doi = {10.47869/tcsj.77.9.5}
}
EUVIMED